Hi, I’m Vaibhav Raina
AI Engineer
AI Engineer
Developer Tools
Developer Tools
Bengaluru, India
Bengaluru, India
I build AI developer tools in Bengaluru: 1,000+ users and $1K MRR on Cheetah AI, my agentic IDE.
About Me
Hi, I’m Vaibhav Raina. I build the tool and context layer that lets AI models do real work in real environments. Cheetah AI exposes IDE and browser operations as tool calls to reasoning models. GrebMCP, my code-search MCP server, serves 300+ users on a BM42 and RRF pipeline.
Hi, I’m Vaibhav Raina. I build the tool and context layer that lets AI models do real work in real environments. Cheetah AI exposes IDE and browser operations as tool calls to reasoning models. GrebMCP, my code-search MCP server, serves 300+ users on a BM42 and RRF pipeline.
Cheetah AI users
Cheetah AI users
Cheetah AI users
Monthly revenue, USD
Monthly revenue, USD
GrebMCP tokens/day
GrebMCP tokens/day
Agents That Act
Context That Fits
Tools People Use
Shipped, Not
Prototyped
Agents That Act
Context That Fits
Tools People Use
Shipped, Not
Prototyped
Projects
Experience
What I Build
Agent Tool Surfaces
I expose IDE and browser operations as native tool calls, so models write, compile and deploy.
Agent Tool Surfaces
I expose IDE and browser operations as native tool calls, so models write, compile and deploy.
MCP Server Engineering
I build Model Context Protocol servers that give coding agents context-aware search over a codebase.
MCP Server Engineering
I build Model Context Protocol servers that give coding agents context-aware search over a codebase.
Retrieval Pipelines
BM42 sparse embeddings with Reciprocal Rank Fusion cut token spend and make agents ~30% faster.
Retrieval Pipelines
BM42 sparse embeddings with Reciprocal Rank Fusion cut token spend and make agents ~30% faster.
Approval-Gated Actions
Macy holds every sensitive action until the user replies, and writes each step to an audit trail.
Approval-Gated Actions
Macy holds every sensitive action until the user replies, and writes each step to an audit trail.
On-Device Inference
I optimize inference to meet the latency and memory limits of Samsung Galaxy hardware.
On-Device Inference
I optimize inference to meet the latency and memory limits of Samsung Galaxy hardware.
Python and TypeScript
I ship Cheetah AI and GrebMCP in Python and TypeScript, with Node.js, Express, MongoDB and Docker.
Python and TypeScript
I ship Cheetah AI and GrebMCP in Python and TypeScript, with Node.js, Express, MongoDB and Docker.
Proof
1,000+ users, $1K MRR
Cheetah AI, cheetahai.co
Cheetah AI is an IDE where reasoning models write, compile, audit and deploy end to end. It runs on a custom Electron layer inside a VS Code fork.
1B+ tokens processed
VS Code extension, 700+ users
The predecessor to Cheetah AI shipped as a VS Code extension. It reached 700+ users and processed 1B+ tokens with no marketing spend.
~30% faster agents
GrebMCP, grebmcp.com
GrebMCP is a context-aware code-search MCP server for coding agents. It carries 300+ users, 50+ of them active, and runs ~500K tokens/day.
3D meshes on device
Samsung R&D Institute India
On-device latency and memory limits on Galaxy hardware shaped the pipeline I built to reconstruct 3D meshes from single-view 2D images.
IEEE CSITSS 2025
Multimodal exam proctoring
A Multimodal Surveillance System for Detecting Cheating Behaviors in Online Exams, written with 4 co-authors, is published by IEEE at CSITSS 2025.
5 hackathon wins
Cloudflare, Residency BLR
Wins include 1st place and a cash prize at the college Innovation Hackathon, the Cloudflare hackathon and the Growth Hackathon by Residency BLR.
5 hackathon wins
Cloudflare, Residency BLR
Wins include 1st place and a cash prize at the college Innovation Hackathon, the Cloudflare hackathon and the Growth Hackathon by Residency BLR.
IEEE CSITSS 2025
Multimodal exam proctoring
A Multimodal Surveillance System for Detecting Cheating Behaviors in Online Exams, written with 4 co-authors, is published by IEEE at CSITSS 2025.
3D meshes on device
Samsung R&D Institute India
On-device latency and memory limits on Galaxy hardware shaped the pipeline I built to reconstruct 3D meshes from single-view 2D images.
~30% faster agents
GrebMCP, grebmcp.com
GrebMCP is a context-aware code-search MCP server for coding agents. It carries 300+ users, 50+ of them active, and runs ~500K tokens/day.
1B+ tokens processed
VS Code extension, 700+ users
The predecessor to Cheetah AI shipped as a VS Code extension. It reached 700+ users and processed 1B+ tokens with no marketing spend.
1,000+ users, $1K MRR
Cheetah AI, cheetahai.co
Cheetah AI is an IDE where reasoning models write, compile, audit and deploy end to end. It runs on a custom Electron layer inside a VS Code fork.
FAQs
Questions recruiters ask
Availability, what I own on each product, how the numbers were measured, and where the research sits.
1.
When are you available for a full-time role?
2.
What does your IEEE CSITSS 2025 paper cover?
3.
What did you personally build on Cheetah AI?
4.
How did you measure GrebMCP’s ~30% speed-up?
5.
Is Macy shipped, or is it still pre-launch?
6.
What did you build at Samsung R&D Institute India?
1.
When are you available for a full-time role?
2.
What does your IEEE CSITSS 2025 paper cover?
3.
What did you personally build on Cheetah AI?
4.
How did you measure GrebMCP’s ~30% speed-up?
5.
Is Macy shipped, or is it still pre-launch?
6.
What did you build at Samsung R&D Institute India?
Contact
Email vaibhavraina12345@gmail.com about roles from 2026 in Bengaluru or remote, or questions about Cheetah AI, GrebMCP and Macy.